using 520 samples train 519 and test 1 using LOOCV

hi i have 3600x520 matrix (b) for patterns and a 26x520 matrix (t) for targer values

i want to train a perceptron using newp(), train() and want to use LOOCV for the testing

520 are samples and 3600 is number of pixels

now im not clear on this part, i think i need to train the net for 519 patterns and test it with the one it leaves out

it will take a long long time so i think im wrong on that

first i need to know how LOOCV works

and how to implement it on this problem

thank you